Xi Jinping and Donald Trump will sit down this week to hash out a plethora of thorny issues from trade to Taiwan. One thing both sides agree on: neither can afford to tap the brakes on AI.
While Silicon Valley debates the possibility of artificial intelligence wiping out humanity, Washington and Beijing appear more worried about global technological supremacy. The two countries have signaled no intention to slow down AI development — a debate that flared anew after Anthropic PBC co-founder Dario Amodei warned of the dangers of unbridled progress in a 3,800-word communique.
Trump almost immediately dismissed calls for moderation as a “hoax” and vowed to maintain American leadership. Beijing sees little reason to slow down in part because it’s already behind the US in AI — though closing the gap — and constrained by American curbs on access to advanced chips by Nvidia Corp.
Just two days before Trump’s set to personally greet Xi at a military airbase, Alibaba Group Holding Ltd. declared it was building a mega model of up to 10 trillion parameters, rolling out AI chips it called China’s most powerful, and planning a 20-gigawatt data center buildout that could translate into $160 billion of cloud revenue, by one estimate.
That capped a frenetic few days for Chinese AI, with Z.AI Co. and Xiaomi Corp. showcasing how their AI systems improve themselves, and Huawei Technologies Co. accelerating the debut of high-powered AI chips intended to replace Nvidia’s processors. Manus – the agentic AI pioneer finally clear of Meta Platforms Inc. – sought to raise capital at a $4 billion valuation to bankroll its next act.
“Neither side wants to give up the development or slow down the development,” said Elly Rostoum, a senior fellow with the tech policy program at the Center for European Policy Analysis and a former US intelligence analyst. “But they understand that this is serious enough that they have to talk, that the companies have to talk.”
The shared desire by the US and China to push forward suggests Thursday’s summit is more likely to produce narrowly defined guardrails rather than speed limits on AI, even as they take tentative steps to build consensus around critical risks. US Treasury Secretary Scott Bessent said Washington proposed an emergency hotline for AI incidents and discussed a dialogue mechanism with his Chinese counterparts at a Sunday meeting in New York.
“China already sees itself as being forced to pace because of US tech containment and chip export controls,” Selina Xu, who leads China and AI policy research in the office of former Google chief Eric Schmidt, said in an online post. “One remark I’ve heard from Chinese policymakers is ‘we’re already being slowed down.’”
See more: Why Big Tech’s AI Capex Is Now Outrunning Cash Flow
Open-Weight Advocate
To close the gap with the US, China has seized on open-weight AI as a strategy to catch up at home and expand its influence abroad. Beijing is keen to keep up that momentum.
Breakout models such as DeepSeek’s R1 and Moonshot’s Kimi K3 can be downloaded and adapted by developers worldwide, allowing Chinese laboratories to build on one another’s work and squeeze more performance from limited computing resources. This has also won fans overseas and helped embed Chinese software in workflows and products, paving the way for monetization.
Moonshot, for example, is building a team of specialized engineers to help clients deploy and customize models, starting in China before expanding globally. It’s also entering commercial partnerships with American cloud providers like Fireworks AI and Baseten to host them.
Z.ai has helped Malaysia build sovereign AI infrastructure, while Saudi Arabia’s state-backed Humain commissioned MiniMax Group Inc. to build a national platform based on one of the Chinese startup’s open models.
Their popularity — a combination of cost, flexibility and performance — has threatened the ability of leading US AI labs to charge for access to their proprietary offerings. Anthropic and OpenAI have both alleged that Chinese companies use the output from their frontier models to create rival systems at a fraction of the cost, a process known as distillation.
The MiniMax deal drew immediate backlash from US national security hawks. Chris McGuire, a senior fellow at the Council on Foreign Relations, wrote on X that Washington was “getting played” after agreeing to sell AI chips to Middle Eastern nations on the promise they would rely on an American software stack — only for Saudi Arabia to adopt Chinese technology.
The risk is that these allegations, which Bessent and other officials have echoed, could lead to tougher enforcement of export controls to starve China of the computing resources required to keep up.
Room for Agreement
The focus on AI risks suggests a shift in emphasis for a rivalry long dominated by US limits on China’s access to advanced chips. Export controls remain a major source of tension and may still emerge during Xi’s visit, but they have received less attention in the days before his arrival.
US Trade Representative Jamieson Greer said export controls were not on the agenda at his Sunday meeting with Chinese Vice Premier He Lifeng and chief trade negotiator Li Chenggang. Beijing hasn’t confirmed the US account of the exchange, saying only that the two sides discussed AI.
China’s continued grip on the global supply of critical minerals likely gives Xi crucial leverage, while Chinese companies including Alibaba accelerate efforts to produce advanced chips that would make the country less vulnerable to Washington’s curbs.
But China is not dismissing the technology’s dangers altogether. Since ChatGPT’s debut in 2022, regulators have introduced rules governing chatbots and proposed restrictions on virtual companions for children. Security officials have also warned about threats of the tools being abused by third parties. This week, the Information reported that China’s cyberspace regulators were investigating potential data breaches by Moonshot and DeepSeek, underscoring Beijing’s intent to police the industry especially around sensitive content.
Those concerns offer potential common ground with the Trump administration, even if they don’t align with the grave risks identified by Amodei and seconded by OpenAI’s Sam Altman and SpaceXAI’s Elon Musk.
“The existential risk from AI is really not a major topic in China, and not a major topic in the Chinese policymaking world,” said Kyle Chan, a fellow with the Brookings Institution. “There’s growing convergence in some of the concerns around other major AI risks, namely around cybersecurity, biological weapons, non-state actors.”
But any agreement on those areas will have to overcome profound distrust about whether the other side is genuinely addressing safety or merely seeking an advantage.
Amodei has argued that preventing China from gaining an AI edge is paramount for safety, a position he’s used to justify tougher export controls. Beijing sees that logic as evidence that calls for safety can serve as a rationale for containment, calling it “fearmongering.”
“Cooperation remains limited for the two countries due to low level of trust,” said George Chen, a partner at consultancy The Asia Group. Beijing continues to believe Washington is seeking to suppress China’s rise in AI and other emerging technologies, a perception that will shape the pace and scope of future engagement, he said.
A message from Advisor Perspectives and VettaFi: Discover something new! Click here to register for our upcoming webcasts.
Bloomberg News provided this article. For more articles like this please visit
bloomberg.com.
Read more articles by
